Implementing Real-time Vision AI Apps Using NVIDIA DeepStream SDK

Discover how to implement real-time vision AI applications using NVIDIA DeepStream SDK in this 45-minute conference talk. Explore how companies worldwide are modernizing infrastructure with IoT sensors for rich insights, from patient fall detection to traffic flow analysis and customer behavior understanding. Dive into the latest capabilities of DeepStream SDK, witness demonstrations of reference apps built with DeepStream SDK, Transfer Learning Toolkit, and Vision AI pre-trained models. Learn about challenges in video analytics, speeding up intelligent video analytics training and inference, and solving industry-specific problems. Explore the DeepStream SDK overview, graph architecture, key features, and Python bindings. Gain insights into IoT applications, secure communication, over-the-air model updates, and native integration with Triton Inference Server. Discover how to deploy open-source models, unlock greater performance, and implement end-to-end AI using Transfer Learning Toolkit and DeepStream. Examine reference apps for people analytics, social distancing, and face mask detection. Access resources to start building high-performance vision AI applications today.

Intro
CHALLENGES WITH VIDEO ANALYTICS
SPEEDUP INTELLIGENT VIDEO ANALYTICS TRAINING & INFERENCE Use NVIDIA Purpose-Built Pretrained Models To Reduce Development Time by 10%
SOLVE INDUSTRY-SPECIFIC PROBLEMS DeepStream provides optimized, deployment ready pipeline with exceptional performance
DEEPSTREAM SDK OVERVIEW Streaming analysis toolkit for based multisensor processing, video and image understanding APPLICATIONS AND SERVICES
DEEPSTREAM GRAPH ARCHITECTURE
DEEPSTREAM 5.0 KEY FEATURES
DEEPSTREAM PYTHON BINDINGS
SAMPLE APPLICATIONS TO GET STARTED TODAY DeepStream Pipelines Using Python
IOT: BIDIRECTIONAL MESSAGING
SECURE COMMUNICATION USING KAFKA Secure edge-to-cloud messaging
IOT: OVER-THE-AIR MODEL UPDATE Eliminate latency be update on the fly
FASTER SEARCHABILITY WITH SMART RECORD
NATIVE INTEGRATION WITH TRITON INFERENCE SERVER Prototype End-to-End System Quickly
TRITON- DEEPSTREAM ARCHITECTURE
DEPLOYING AN OPEN SOURCE MODEL
UNLOCK GREATER PERFORMANCE USING DEEPSTREAM
END-TO-END AI USING TLT AND DEEPSTREAM
TLT PURPOSE-BUILT MODELS DEPLOYED USING DEEPSTREAM Achieving Higher Accuracy & Real-Time Performance
END-TO-END REFERENCE APPS Get started today
PEOPLE ANALYTICS FOR CROWDED SPACES Reference architecture demo
SOCIAL DISTANCING APP Reference architecture demo
FACE MASK/NO MASK DETECTION Reference architecture demo
DEEPSTREAM WORLDWIDE ADOPTION
SUMMARY
RESOURCES Get started today